CtrlA: Adaptive Retrieval-Augmented Generation via Inherent Control
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866910631833108480 |
|---|---|
| author | Liu, Huanshuo Zhang, Hao Guo, Zhijiang Wang, Jing Dong, Kuicai Li, Xiangyang Lee, Yi Quan Zhang, Cong Liu, Yong |
| author_facet | Liu, Huanshuo Zhang, Hao Guo, Zhijiang Wang, Jing Dong, Kuicai Li, Xiangyang Lee, Yi Quan Zhang, Cong Liu, Yong |
| contents | Retrieval-augmented generation (RAG) has emerged as a promising solution for mitigating hallucinations of large language models (LLMs) with retrieved external knowledge. Adaptive RAG enhances this approach by enabling dynamic retrieval during generation, activating retrieval only when the query exceeds LLM's internal knowledge. Existing methods primarily focus on detecting LLM's confidence via statistical uncertainty. Instead, we present the first attempts to solve adaptive RAG from a representation perspective and develop an inherent control-based framework, termed \name. Specifically, we extract the features that represent the honesty and confidence directions of LLM and adopt them to control LLM behavior and guide retrieval timing decisions. We also design a simple yet effective query formulation strategy to support adaptive retrieval. Experiments show that \name is superior to existing adaptive RAG methods on a diverse set of tasks, the honesty steering can effectively make LLMs more honest and confidence monitoring is a promising indicator of retrieval trigger.Our code is available at \url{https://github.com/HSLiu-Initial/CtrlA}. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2405_18727 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | CtrlA: Adaptive Retrieval-Augmented Generation via Inherent Control Liu, Huanshuo Zhang, Hao Guo, Zhijiang Wang, Jing Dong, Kuicai Li, Xiangyang Lee, Yi Quan Zhang, Cong Liu, Yong Computation and Language Artificial Intelligence Information Retrieval Retrieval-augmented generation (RAG) has emerged as a promising solution for mitigating hallucinations of large language models (LLMs) with retrieved external knowledge. Adaptive RAG enhances this approach by enabling dynamic retrieval during generation, activating retrieval only when the query exceeds LLM's internal knowledge. Existing methods primarily focus on detecting LLM's confidence via statistical uncertainty. Instead, we present the first attempts to solve adaptive RAG from a representation perspective and develop an inherent control-based framework, termed \name. Specifically, we extract the features that represent the honesty and confidence directions of LLM and adopt them to control LLM behavior and guide retrieval timing decisions. We also design a simple yet effective query formulation strategy to support adaptive retrieval. Experiments show that \name is superior to existing adaptive RAG methods on a diverse set of tasks, the honesty steering can effectively make LLMs more honest and confidence monitoring is a promising indicator of retrieval trigger.Our code is available at \url{https://github.com/HSLiu-Initial/CtrlA}. |
| title | CtrlA: Adaptive Retrieval-Augmented Generation via Inherent Control |
| topic | Computation and Language Artificial Intelligence Information Retrieval |
| url | https://arxiv.org/abs/2405.18727 |